Mar 27, 1990 - The EMBO Journal vol.9 no.7 pp.21 59 - 21 67, 1 990. The 3-globin ... We have shown previously that a 1.9 kb fragment comprising HSS 2.
The EMBO Journal vol.9 no.7 pp.21 59 - 21 67, 1 990
The 3-globin dominant control region: hypersensitive site 2
Sjaak Philipsen, Dale Talbot, Peter Fraser and Frank Grosveld Laboratory of Gene Structure and Expression, National Institute for Medical Research, The Ridgeway, Mill Hill, London NW7 1AA, UK Communicated by F.Grosveld
The Dominant Control Region (9CR) of the human (globin gene locus consists of four strong hypersensitive sites (HSS) upstream of the e-globin gene. Addition of these sites confers copy number dependent expression on the human ,B-globin gene in murine erythroleukaemia cells and transgenic mice, at levels comparable with the endogenous mouse globin genes. We have shown previously that a 1.9 kb fragment comprising HSS 2 accounts for 40-50% of the full effect of the DCR. In this paper we describe a deletional analysis of HSS 2. We show that a 225 bp fragment is sufficient to direct high levels of expression of the human ,B-globin gene which is copy number dependent and integration site independent. This 225 bp fragment overlaps the major region that is hypersensitive 'in vivo'. DNase I footprinting shows the presence of four binding sites for the erythroid specific protein NF-E1; the three other footprinted regions display a remarkable redundancy of the sequence GGTGG and bind a number of proteins including Spl and the CACC box protein. The significance of these results for the regulation of globin gene expression is discussed. Key words: Dominant Control Region/erythroid factors/0globin
Introduction The human ,B-globin gene locus consists in the 5' to 3' direction of the e-globin gene, which is expressed in embryonic stages, the G-y and A-y genes, which are expressed during fetal development and the b- and (3-globin genes which are expressed during adult life. The gene cluster comprises 60 kb (for review, see Collins and Weissman, 1984) and expression is restricted to cells of the erythroid lineage. To understand the mechanism underlying the stage and tissue specific expression of the genes in the human f-globin gene cluster, DNA fragments containing the individual genes were used to generate transgenic mice (Magram et al., 1985; Chada et al., 1985; Kollias et al., 1987; Behringer et al., 1987). It was shown that the Py-globin genes behaved like the mouse embryonic ,BH1 gene, while the human ,B-globin gene followed the expression pattern of the mouse 3major gene. However, expression levels were very low and dependent on the integration site in the mouse genomic DNA. Recently, we have shown that the addition of a region containing four very strong hypersensitive sites (HSS) found upstream of the c-gene (Tuan et al., 1985; Forrester et al., Oxford University Press
1987) to a ,3-globin gene fragment containing all its known local regulatory elements overcomes the dominant action of position effects (Grosveld et al., 1987). Moreover, expression of the transgene was found to match that of the endogenous globin genes and was dependent on the copy number of the integrated construct. For this reason, the construct was termed the 'minilocus' and the HSS were called the Dominant Control Region (DCR) of the human ,B-globin gene cluster. The four 5' HSS were contained in a 20 kb fragment in the original minilocus; we subsequently linked the HSS together as a 6.5 kb fragment and showed that this configuration, designated the microlocus, was comparable with the minilocus as defined in murine erythroleukaemia (MEL) cells and transgenic mice (Blom van Assendelft et al., 1989; Talbot et al., 1989). The analysis of a similar construction has recently been reported by Forrester et al. (1989). The discovery of DCR sequences in the human 13-globin gene locus has opened the way for a realistic approach towards gene therapy and to the development of animal models for human haemoglobinopathies. The feasibility of the latter has been demonstrated in our laboratory by the development of transgenic mice that carry the human allele for sickle haemoglobin under the control of the DCR (Greaves et al., 1990). The erythrocytes of a mouse producing more f3s than endogenous mouse fi-globin sickled both in vivo and in vitro, providing a mouse model to study anti-sickling drugs and gene therapy protocols. For somatic gene therapy, introduction of the transgene via retroviral insertion in stem cells appears to be the most realistic approach at the moment. For this, the development of retroviral constructs passaging with high titres is essential to obtain a high frequency of infection in the targeted cells. When single HSS of the DCR were tested in MEL cells, it was shown that HSS 1 and 4 gave a level of 10% of the microlocus, while both HSS 2 and 3 conferred -50% to a linked 03-globin gene (Collis et al., 1990). These results have been confirmed in transgenic mice (Fraser et al., 1990). In this paper we describe a detailed analysis of HSS 2. Fine mapping shows that the HSS 2 is, in fact, a small hypersensitive region rather than a defined site and functional analysis of MEL cells and transgenic mice shows that a core 225 bp fragment, coinciding with the HSS, allows high level of position independent expression. -
Results Fine mapping of HSS 2 in MEL cells
The locations of the HSS in the ,B-globin DCR were previously mapped on large restriction fragments (Tuan et al., 1985; Forrester et al., 1987; Grosveld et al., 1987). This allowed us to construct a smaller fully functional DCR containing the individual HSS on 1-2 kb fragments (Talbot
2159
S.Philipsen et al.
et al., 1989; Figure IC). In this microlocus construct, HSS 2 is present on a 1.9 kb HindHI fragment. To determine the position of HSS 2 more accurately, we took advantage of a MEL cell line containing four copies of the minilocus construct (clone C, Blom van Assendelft et al., 1989). Nuclei were isolated (Gorski et al., 1986) and treated with different amounts of DNase I for 5 min on ice. As a nonerythroid control, mouse L-cells containing the same construct were used (Blom van Assendelft et al., 1989). Southern blots of HinduI digested DNA were probed for
4444 £ A
0
10
20
30
af
40
tpi
A
50
60
70
s0
00
B
C
D silI
0SS 2
-
Fig. 1. The Dominant Control Region of the human ,B-globin gene cluster. (A) The human f3-globin gene cluster on the short arm of chromosome 11. The DCR, characterized by four hypersensitive sites 5' to the E-globin, and one hypersensitive site 3' to the fl-globin gene is indicated by vertical arrows: (B) The minilocus described by Grosveld et al. (1987), combining the DCR and the 3' hypersensitive site with the human ,B-globin gene. (C) The microlocus constructed by Talbot et al. (1989), containing the DCR region as four restriction fragments of 1-2 kb. (D) The plasmid constructs used in this study. All the fragments tested are cloned in the HpaI site at 800 bp in front of the cap site of the human (3-globin gene. The Sall and EcoRV sites used to isolate the DNA containing the gene plus HSS 2 for transgenic mice are indicated.
HSS 2 via indirect end-labelling. With a probe specific for the 5' end of the 1.9 kb HindIll fragment, we reproducibly found two hypersensitive regions in MEL cells, the 5' region being the weakest (Figure 2, panel B). With the 3' probe, only the stronger 3' region is seen (Figure 2, panel A). We were unable to detect hypersensitivity in L-cells (Figure 2, panels A and B). From these results we infer that the 5' hypersensitive site is located between nucleotides 950 and 1150, and the 3' site between 1250 and 1550 (Figure 3, top line). Functional analysis of HSS 2 deletions A series of HSS 2 deletions was made (Figure 3) and cloned in the HpaI site at position -800 in front of the human fglobin gene as shown in Figure 1D. Except when indicated otherwise, the natural sense orientation was used. Plasmids were linearized with PvuI and transfected into MEL cells by electroporation. After selection in G418 containing medium, populations were induced by the presence of 2% DMSO for 4 days and expression of the construct was measured by quantitative SI analysis using expression of the endogenous mouse a-globin genes as a control (Figure 4). Construct 1 is the 1.9 kb HindIII fragment, which serves as a reference for the full activity of HSS 2 on its own. Construct 2 is a doublet of this fragment. Interestingly, expression of the test gene is remarkably higher with this doublet, indicating that this tandem array allows cooperativity between the two HSS 2 fragments (Table I). Constructs 3-8 are various 5' and 3' fragments of HSS 2; analysis of their expression patterns allows us to draw the conclusion that only the fragments containing the major 3' HSS (Figure 3, Table I) give the full expression observed with the original HSS 2 fragment. Constructs 6 and 7 are particularly instructive, since construct 6 contains the 5' part of the 1.9 kb HindmI fragment, including the weak 5' HSS, while construct 7 contains the remaining 3' part comprising the 3' HSS. The 5' border of construct 7 is just inside the 5' border of the strong HSS as mapped in MEL cells (Figures 2 and 3). Only construct 7 gives a level of
Fig. 2. Fine mapping of HSS 2 in MEL and L-cells. Nuclei were isolated and treated with DNase I (see Materials and methods). DNA was purified and restricted with HindIII and Southern blotted. MEL DNA was from uninduced (-) cells or from cells induced for 2 days with 2% DMS0 (+). The amount of DNase I (ug/mg DNA) is indicated. The major and minor hypersensitive regions are indicated by dashed blocks. (A) Hybridized with the 3' BstNI-HindIII fragment. (B) Hybridized with the 5' HindIII-BamHI fragment. The position of the major and the minor hypersensitive regions within the 1.9 kb fragment are indicated in Figure 3. 2160
j3-globin dominant control region
expression comparable with the full site; the level off construct 6 is essentially the same as that of the vectorr without any DCR fragment (Table I). These results prompted us to test a 225 bp HphI -Fnu4HI1 fragment which is slighfly smaller than the major HSS. This3 fragment was tested in both the sense (construct 9) and the antisense (construct 10) orientation. The results are shownI in Figure 4, panels A and B and Table I. Even this smallI restriction fragment provides nearly full expression whenI compared with constructs 1,3,5,7 and 8, independent off orientation. This predicts that this small fragment is capable of providing position independent high levels of expression. However, analyses using MEL cell populations are rather limiting in assaying position independent expression. First x
I
of all, cell transformants would have to be cloned and analysed for expression and transgene integrity. Second, the test construct must be integrated into active chromatin regions since the transfected cells are selected in G418 for expression of the linked marker tk-neo gene, thus biasing the result. Finally, only a limited copy number range can be achieved in these MEL cells and transcription levels per gene are decreased with increasing copy numbers (>4-5) with similar type small constructs (Talbot et al., 1990).
I I
0
1, 2
1000
500
1 HindIII
XmnI
1500
I
I
SauI
HphI
(k
5
6'C 7
Ist
Aa
satIII
AvaII
9,10
HphI
Fnu4HI
Fig. 3. Deletions of HSS 2 tested in stably transformed MEL cells. All fragments are cloned in the HpaI site of the vector GSE 1273 in the sense orientation, with the exception of construct 10, which is the antisense orientation of construct 9. Construct 2 contains the 1.9 kb HindlIl fragment duplicated in a tandem array. Hatched boxes indicate hypersensitive sites.
Transgenic mice To test for position independent levels of expression we used the Sall-EcoRV fragment of construct 9 (Figures 1 and 3) for microinjection into fertilized mouse eggs. Transgenic 13.5 day fetuses were collected and expression was compared with 13.5 day fetuses containing the 1.9 kb Hindm fragment driving the human f-globin gene (Fraser et al., 1990). To determine copy numbers of seven transgenic fetuses, the initial blots of placenta DNA were probed with a human ,3-globin probe and an endogenous mouse Thy-1 probe as a loading control. To screen for mosaicism of the transgene, DNA from body, head and yolk sac was then also analysed, using the same probes. This showed that mice 31 and 32 were mosaic, since different tissues had a different ratio of human j-globin to Thy-i signal on the blots (not shown). Figure 4B and Table I show the results of a quantitative RNA analysis in fetal livers of the seven transgenic fetuses. As expected for mosaic mice, 31 and 32 show low levels of expression per copy of the gene. The other mice show expression levels per gene of 40-50% when compared with the construct containing the entire HSS 2 region (nos. 49 and 59). The exceptions are the mice with extremely high copy numbers (nos. 30 and 354), showing lower levels of
A
Fig. 4. SI analysis of the hypersensitive site 2 deletions. In panel A a probe for the 3' end of the human ,B-globin mRNA was used, in panel B a probe for the 5' end. The probe for the endogenous mouse ce-globin mRNA was for the 5' end in all cases. The combination of a human 3' ,Bglobin probe and the mouse 5' co-globin probe results in a background (3-signal which was subtracted from all (3 signals to obtain the values in Table I. Constructs are numbered as in Figure 3. Transgenic mice 1, 20 and 29-33 were made with the Sail-EcoRV fragment of construct 9 (Figures 1 and 3). Mouse 36 is the non-transgenic control. Transgenic mice 49, 59 and 354 contain the 1.9 kb HindIH fragment 3' to the 4.8 kb Bgll fragment of the human ,B-globin gene (Fraser et al., 1990). Specific activities: panel A, left: 3:ca = 1:5.3; panel A, right: (3:ac = 2.7:1; panel B, ,B: x = 1.5:1.
2161
S.Philipsen et al. Table I. Expression of hypersensitive site 2 constructs in MEL cells and transgenic mice MEL populations
Construct
HO3/Mcx
Copy no. Mouse ca Human ,B %
IA
>8 1 2 2 2.5 2 2 3 2 4A 1 4B 1 4C 1 SA 3 SB 2.5 SC 2 6A 1 6B 0.5 6C 0.5 6D 3 6E dl 7A 3 7B 2 7C 2 8A 4.5 8B 3 8C 4 8D dl 8E 3 8F 3 9A 5 9B 2 9C 3.5 lOA 1.5 lOB 7.5 lOC 2.0 1273A no DCR 1 1273B no DCR 1.5 1273C no DCR 2 1401A ul 3.5 1401B Al 4.5 5 1401C ytl
lB IC 2A 2B 2C 3A 3B 3C
13282 14785 17882 8687 17599 13672 8733 7075 7174 11304 15832 9121 6597 10904 17914 17603 16339 10372 32389 nd 10866 20820 15761 8769 7134 8234 11367 10617 9190 9222 11403 12922 9025 9688 10754 13951 20703 17178 13210 12606 12999
3509 1968 3127 4636 10770 7767 2177 3345 2266 182 134 47 2822 3711 4113 59 81 44 122 nd 3411 4171 3145 5046 2285 4060 1537 3257 4555 4634 3631 4543 2369 5436 3052 117 314 346 12716 14182 13088
24 13 18 53 61 57 25 47 32 2 1 0.5 43 34 23 0.3 0.5 0.4 0.4 31 20 20 58 32 49 13 30 50 50 32 35 26 56 29 1 1.5 2 96 112 101
% exp/copy Av. 52 36 106 98 114 50 63 64 8 4 2 57 54 46 1 2 3 0.53
44
42 40 40 52 43 49 40 67 40 64 40 69 30 58 4 6 4 110 100 80
Transgenic mice Mouse no.
Copy no. Mouse a Human , % H,3/Ma % exp/copy
I 20 29 30 31 mosaic 32 mosaic 33 354
11 5 1
59
50 30 13 19
50
5 49 2 36 non transgenic 0
8101 27494 24666 1814 11017 2874 715 7538 8103 6655 3228
11060 13696 3084 4696 4171 1201 1388 14504 8536 3492 43
136 49 12 259 38 42 194 192 105 53 1
49 39 48 21 5 13 41
15 84 106 -
Bands were cut out of the gel and Cerenkov counted. A similar sized gel fragment just above the band of interest was also counted for background correction. The data given are corrected for the relative specific activities of the probes used. Copy numbers were determined from Southern blots as described (Talbot et al., 1990). The microlocus controls were those used by Talbot et al., (1989). dl, deletion; y1, microlocus; nd, not determined.
2162
expression, a phenomenon previously observed for both HSS 2 and 3 (Fraser et al., 1990). This indicates that although the small HphI -Fnu4HI fragment shows a reduced activity when compared with the large Hindm fragment in fetal liver versus MEL cells, this core fragment has retained the capacity to provide copy number dependent, integration site independent expression on the f3-globin gene. DNase I footprinting Based on the expression data in transgenic mice and MEL cells we analysed protein-DNA interactions in the 225 bp HphI-Fnu4HI fragment by in vitro DNase I footprinting. Figure 5 shows the results obtained with nuclear extracts (Gorski et al., 1986) from uninduced MEL cells, representing the non-expressing adult erythroid stage and anaemic adult spleen expressing the globin genes at high levels. Apart from one hypersensitive site which is stronger with the anaemic spleen extract (indicated by the top arrow in footprint 2 in Figure SA), the same patterns are observed with both extracts, i.e. six footprints, numbered 1-6 in Figure 5 and summarized in Figure 6. Footprinting with extracts from fetal liver, induced MEL cells, K562 cells, HeLa cells and adult liver showed that the non-erythroid tissues lacked footprints 1, 3 and 5. No differences were observed between the footprints obtained with the different erythroid cell extracts. The three erythroid specific footprints 1, 3 and 5, represent binding sites for the major erythroid specific factor NF-E1 (Wall et al., 1988; Evans et al., 1988; Tsai et al., 1989). As a hallmark of NF-El binding very strong hypersensitive sites are observed immediately upstream of the binding site (indicated by horizontal arrows in Figure 5 and vertical arrows in Figure 6). Competition experiments with NF-E1 specific oligonucleotides of the human (3-globin gene 3' enhancer (Wall et al., 1988) also confirm that footprints 1, 3 and 5 are NF-El sites (not shown). Another consensus NF-El site is present in footprint 4, but the presence of NFEl is obscured by protein binding to the neighbouring remarkable sequence motif that is repeated in footprints 2, 4 and 6. This motif 5 '-GnnnGGTGG-3' occurs in the same orientation twice in footprint 2, three times in footprint 4 and once in footprint 6 (Figure 6). The presence of additional Gs in the sequence predicts that at least part of these footprints is generated by the general transcription factor Spl (Kadonaga et al., 1987) and the CACC binding protein TEF2 (Xiao et al., 1987). We therefore carried out bandshift/competition experiments to determine which complexes are formed with these regions and compared these with genuine Spl or CACC box binding sites. Figure 7 shows that each of the oligonucleotides forms a number of complexes in MEL extracts including Spl and TEF2, which are also present in HeLa cells (not shown). The Spl oligonucleotide (Figure 7, panel Spl) contains a dimer Spl binding site (Gidoni et al., 1985) which forms four complexes, the doublet due to binding of the 95 kd and 105 kd forms of Spl (Jackson and Tjian, 1988) and a slower mobility complex [labelled (2 x)] which is the result of two Spl binding sites on the same oligonucleotide. The nature of the faster mobility complex (also labelled Spl) which is observed in all of our extracts is, at present, not clear. It could be a degradation product of Spl (Gustafson and Kedes, 1989), but could also be a different protein (Xiao et al.,
3-globin dominant control region A
,,)
~
~
~
~
::
urn
4
-
::
41k 4,
.
.4,
a:
B
5
* 4,
4,
4
-4--
.....
4,
4w U, a
;I&
Or-S
4,
U,
a. 1.
Fig. 5. DNase I footprinting of the 225 bp HphI-Fnu4HI fragment. The DNA in the 'no protein' (-) lanes was treated with 1, 0.5, 0.25 and 0.125 ,ug DNase I from left to right in each panel. DNA pre-incubated with 10 ,tg nuclear extract from MEL cells or anaemic spleens (an.spl.) was treated with 1 ug DNase I, DNA pre-incubated with 50 itg nuclear extract was treated with 2 jg DNase I. Panel A shows the sense strand, panel B the antisense strand. Footprinted regions are indicated by numbered brackets, a weak footprint by a dotted bracket. Arrows indicate hypersensitive sites.
1987). In support of the latter possibility, is the fact that an antibody specific for Spl (gift of S.Jackson) does not affect the mobility of this band (E.Spanopoulou and F.Grosveld, unpublished). The CACC box oligonucleotide (derived from the f-globin gene promoter, Figure 7, lanes CACC) binds the CACC box protein (labelled CACC) in addition to those proteins bound by the Spl oligonucleotide. It also binds a number of fast mobility complexes. The site 2 probes 2, 4 and 6 (see Figure 6) specifically form a number of additional complexes (labelled 1-10). Of the additional complexes, only number 1 can be competed efficiently by the CACC box oligonucleotide, while none are competed by the Spl oligonucleotide. It is at present not clear which of these proteins is functionally important.
3
2
1
I
+
4s
;
. .. . . . . . . . . . . . .
TGGTGTGCCAGATGTGTCTATCAGAGGTTCCAGGGAGGGTGGGGTGGGGTCAGGGCTGGCCACCAGCTATCAGGGCCCAG
IlI
ATGGGTTATAGGCTGGCAGGCTCAGATAGGTGGTTAGGTCAGGTTGGTGGT(
t
t e-t
I
t
6
5
1
1~~~~~~~~~~~~~~~~~~~~
:AGGAGAGATAGACZATGA5TAGAGAGCAGACAT3GGAAAGGTGGGGGAGGCACAGCATAGCAGC 1.
Fig. 6. Summary of the footprinted regions in the 225 bp HphI-Fnu4HI fragment. Protected regions on the sense and antisense strand are indicated by brackets, a dotted bracket indicates weak protection; vertical arrows indicate hypersensitive sites. Thin horizontal arrows indicate NF-E1 consensus binding sites; thick horizontal arrows indicate the GGTGG motif.
Discussion The DCR of the human ,B-globin gene cluster has been defined to the DNA region between 5 and 25 kb upstream of the E-globin gene (Grosveld et al., 1987). This region contains four erythroid specific 'super' hypersensitive sites for DNase I (Tuan et al., 1985; Forrester et al., 1987), which were shown to be the functional elements of the DCR (Talbot et al., 1989; Forrester et al., 1989). Deletional analysis of the microlocus construct demonstrated that HSS 2 and 3, as single sites, conferred high levels of expression on the human ,B-globin test gene in a stable transfection assay
(Collis et al., 1990). Interestingly, when tested in transient expression assays in K562 and MEL cells, only HSS 3 was found to stimulate CAT activity (Tuan et al., 1989 and C.H.Chang and P.Dierks, submitted, respectively). This suggests that integration into chromatin is an important prerequisite for proper functioning of the DCR. It clearly distinguishes 'DCR' type elements from classical enhancers, which were originally defined in transient transcription assays (for review see Serfling et al., 1985; Maniatis et al., 1987; Dynan, 1989) and do not necessarily provide integration 2163
S.Philipsen et al.
~~ -9e
Fig. 7. Gel mobility shift and competition assays of the footprinted regions 2, 4 and 6. Probes covering footprints 2, 4 and 6 (Figure 6) and probes for the SV40 GC-box (Spl) and the CACC box of the human ,3-globin gene (CACC) (see Materials and methods) were used in a gel shift assay with nuclear extracts from MEL cells. Competitors were added in 100-fold molar excess as indicated. The position of bands specific for the Spl and CACC proteins are indicated by Spl and CACC labels. Other specific complexes are numbered relative to their mobility.
position independent expression. HSS 3 appears to be a combination of a DCR element (Forrester et al., 1989; Collis et al., 1990; Fraser et al., 1990; Talbot et al., 1990) and a powerful enhancer (Curtin et al., 1989; Tuan et al., 1989; Ryan et al., 1989) stimulating transcription in both stable and transient assays and in transgenic mice. To understand the mechanism by which 'DCR' type elements act, it is therefore important to define the minimal requirements for this activity, as we describe in this paper for HSS 2 of the human ,-globin gene cluster. Precise mapping of HSS 2 reveals that there is a major hypersensitive region of 250 bp in the 3' end of the 1.9 kb Hindml fragment, and a minor site of 150 bp 5' to the major site. The minor site is found only when probing from the 5' end of the HindIm fragment and footprinting analysis (not shown) shows very few, if any, protein binding sites. Since deletional analysis in MEL cells shows that all the constructs containing the major site (1, 3, 5, 7, 8, 9 and 10) retain the full activity of HSS 2, we concentrated our analysis on this part of HSS 2. HSS 1 and 3 have also been mapped in detail in our laboratory (O.Hanscombe, unpublished; Talbot et al., 1990) and a remarkable common feature is that all the sites have core fragments of between 200 and 300 bp in length. This may indicate that there is a size requirement for DCR elements, roughly coinciding with the length of the nucleosomal repeat unit. Based on the mapping results in MEL cells (Figure 3 and Table I), we tested the smallest fragment for HSS 2 in transgenic mice. This fragment has retained the most important properties of the entire 1.9 kb HSS 2, i.e. orientation and integration position independent expression of the fl-globin gene, albeit at a reduced level. The latter could be due to the fact that the smallest fragment has lost some enhancing sequences, e.g. a potential NF-E2 binding site (Mignotte et al., 1989a,b), which is located just upstream of this fragment and could act as an enhanson (Talbot et al., 1990). It also highlights the occurrence of differences between MEL cell assays and transgenic mice analysis, something we have previously observed with HSS 1 (Collis et al., 1990; Fraser et al., 1990). This could be due to stage specific differences between the two systems or some inherent limitation of the MEL cells. DNase I footprinting with erythroid nuclear extracts shows that there are six protected regions in the 225 bp fragment. These comprise at least three NF-El binding sites (Wall et al., 1988; Tsai et al., 1989) and three regions which have -
2164
the motif GnnnGGTGG in common (Figure 6). The NF-El and GGTGG motifs occur in an alternating order and we have noticed that the combination of these motifs is a common feature of many promoter and regulatory elements of erythroid specific genes. The spacing between the two motifs is often 30 bp. Table II lists a number of such combinations to substantiate this observation. We propose that the GGTGG array is a key determinant for gene expression in the erythroid lineage and it will be important to determine which protein interacts functionally with the G-motif in erythroid cells. Spl (Gidoni et al., 1985) and the CACC box binding protein, TEF2 (Xiao et al., 1987) are both abundant in erythroid cells and both bind to this sequence in vitro (Figure 7), but these proteins are not erythroid specific. It has been shown that both of these can interact with other distal transcription elements and their factors to mediate their effect to the transcriptional machinery (Schiile et al., 1988). However, the GGTGG motif is different from the Spl and the CACC consensus and it is clear that a number of other ubiquitous proteins are bound to this region which may provide the main activating function in vivo; a clue to this is provided by the conserved G residues 5' and 3' to the GGTGG repeats in HSS 2 (Table II). Point mutations at all the positions should resolve which of the proteins is active in a functional analysis. It is conceivable that the combination of such a factor and a tissue specific factor (NF-E1) is sufficient to activate and direct high levels of erythroid specific gene expression. Although the NF-El protein has the ability to trans-activate (Tsai et al., 1989; Evans and Felsenfeld, 1989), the presence of NF-E1 sites alone is not sufficient for activation in vivo, since each of the globin genes (without a DCR) contains multiple NF-El binding sites, but is only expressed at low levels and in a highly position dependent manner. Therefore, NF-E1 could be a major activator only in combination with a second factor, where one protein could be the first to bind, without activation of the genes, but enabling the second to bind, thus mediating a transcriptional effect. Finally, the work described here contributes to the construction of a fully functional DCR of the human 3-globin gene cluster containing at least three of the HSS on an -1 kb fragment. This should greatly facilitate the construction of retroviral vectors which passage with titres high enough to transduce cultured bone marrow stem cells efficiently. It would be a significant step forward if fl-thalassaemia could be cured by grafting of autologous 'repaired' bone marrow -
,B-globin dominant control region Table II. Comparison of different GGTGG arrays in erythroid promoters and regulatory elements
AACCTCTGATAGACAC
a 19 s
GGGGAGC-IMZGGTGGG
s 30 a
GGGCCf,T.ATAQCTGG
HSS 2, 1324
AACCTCTGATAGACAC
a 24 s
GGGTGGCIKTCGGTCAG
s 25 a
GGGCCTaGTA,GCTGG
HSS 2, 1329
GGGCCCAGIAGTTA
s 30 s
CAGATAGGTG2TTAGGr
GCTCTCGATIAGGTGG
s 22 s
CAGGTTGGGTGCTGG
s 41 s
CAGGAGAG&LIGACCA
HSS 2, 1412
GCTCTLCAGTAGGTGG
s 32 s
GTGCTGTQCG2AGTCCA
s 31 s
CAGGAGAG&TAGACCA
HSS 2, 1422
CAGGAGAGATAGACCA
s 34 s
GGGAAUAj2IGTGGGAGG
HSS 2, 1487
ATCGTGAGATAGACGT
a 27 s
AGAAGGCTIQQACTCCA
HSS 1, 899
CAGGGXAGAT-QQCAAA
s 26 a
TAGTCAGGTGGTCAGCT
HSS 3, 1019
CAGGGC1LiAT9CAAA
s 42 a
GTTTGAQQTQAGTTTT
HSS 3, 1035
HSS 2, 1395
TGCCATGGTGGTTTGCT
s 27 a
TAATGTIAGTGACGGG
HSS 4, 247
GTTG7CGa99GTGGGGCT
a 24 s
AGTGTGTGATGTTCCC
HSS 4, 247
CAGCAGTGATGGATGG
a 30 a
CACAGG27Tl2AGTCAG
H. £, -165
GCATTGAGATAGTGTG
a 44 a
CCCATGGGTGGAGTTTA
H. Gy, -143
GGCCTATGATAGGGTA
a 10 a
ATTTGGGGTGGGGCCTA
a 40 s
TGTTTAAGATTAGCAT
H. P-gl.enh., +231
TTGTGGGGTGGCGCGTG
a 30 a
GGCTCCAGATTCAGAG
H. a-gl., -713
CGAGCGGGATGGGCGG
s 23 s
GTGGCG,GGTGAGGGTG
H. ax-gl., -202
CGAGCGGGATGGGCGG
s 30 s
GTGGAGGGTGGAGACGT
H. ax-gl., -195
GGAA,GGTGGCCTGG
s 12 s
GGCCTGGGATAACAGC
H. Gph. A, -51
AGGAAGGGTGGGGCCTG
a 31 a
GTAAAGAGATAAGGCC
H. PBGD, -100
GCTGGGTGTGCCC
s 32 s
CCTCAGATAAGACC
CAGCTGGGTGGGGGCAG
s 16 s
GGTTGCAGATAAACAT
Ch.
CAGCTGGGTGGGGGCAG
s 31 a
AAGTCTTGATAGCAAA
Ch. j3A enh., +1882
Rat PK, -51
PiA
enh., +1882
GGTGG G
8
6 14
6
7 14 21 21
A
3
5
1
6
6
5
0
T
3
7
3
5
6
2
C
5
1
3
4
2
Consensus:
-
-
G
-
-
0 21 20
9 15
4
7
6 10
0 0 0 0
7
0
3
1
3
4
0
0 21
4 2
8
4
6
6
0
0
0 0 0 0
1
4
6
9
5
0
G
G
G
r
G
y Cg
T
0
G
1
G
-
r
The GGTGG and NF-El motifs are underlined. Frequencies of nucleotides in each position are given. Note that all the sequences are given in the orientation that allows alignment, as indicated by s (sense) and a (antisense). Distances are calculated from the central base in the GGTGG motif to the central base in the NF-E1 motif RNGATNR. Numbering given for the HSS of the DCR is from the 5' restriction sites used to construct the microlocus (Talbot et al., 1989); others are relative to the cap site. H., human; gl., globin; enh., enhancer; Gph., glycophorin; PBGD, phosphobilinogen deaminase; PK, pyruvate kinase; Ch., chicken.
stem cells, opening the road to the treatment of human
haematopoietic diseases by somatic gene therapy protocols.
Materials and methods Plasmid constructions Plasmid GSE1273 contains the human ,B-globin gene as a 4.8 kb BglII fragment linked to a tk-neod gene (Talbot et al., 1989). All the fragments
tested were blunted and cloned in HpaI digested GSE1273, replacing tht
650 bp HpaI Ifragment in the 5' flanking region of the human ,B-globin gen ie (C:A' ksee FioV11rP rigure i).8 DNase I sensitivity Nuclei were isolated according to Gorski et al. (1986) and resuspended at a DNA concentration of 1 mg/ml in 15 mM Tris pH 7.5, 60 mM KCl, 15 mM NaCl, 0.2 mM EDTA, 0.2 mM EGTA, 0.15 mM spermine, 0.5 mM spermidine. DNase I tWorthington) was added to concentrations
2165
S.Philipsen et al. of 1-8 Ag/ml, and the reaction was started by the addition of MgCl2 to 10 mM and CaC12 to 1 mM final concentration. The reaction was allowed to proceed for 5 min on ice, and stopped by the addition of 1 vol 50 mM EDTA, 1% SDS. DNA was isolated, digested with appropriate restriction enzymes, and analysed by Southern blotting. Tissue culture and cell transfections The MEL cell clone (8-C and the L-cell population 4, each containing 4-5 copies of the human jB-globin minilocus construct, have been described previously (Blom van Assendelft et al., 1989). The MEL cell line C88 was maintained in standard caMEM plus 10% fetal calf serum. Plasmid constructs were linearized at the PvuI site and transfected by electroporation as follows: log phase MEL cells (2 x 107 cells per transfection) were washed and resuspended in 1.5 ml HEPES buffered saline containing 50 itg of the linearized plasmid. After incubation on ice for 10 min the cells were electroporated with three pulses from a Hoefer 'Pro-Genitor' apparatus set to deliver 250 V for 10 ms. After 5 min at room temperature, they were divided to generate three independent transfected populations (Antoniou et al., 1988). MEL cells were induced to undergo erythroid differentiation by incubation in the presence of 2% (v/v) DMSO for 4 days.
Transgenic mice Plasmids were cut with EcoRV and Sall and the fragment containing the human ,B-globin gene was separated from plasmid sequences by agarose gel electrophoresis and recovered from the gel via isotachophoresis (Ofverstedt et al., 1984). DNA was dissolved at 1-2 itg/mi in microinjection buffer and oocytes were injected into the male pronucleus, as described (Kollias et al., 1987). Fetuses were collected 13.5 days after transfer to pseudopregnant foster mothers and analysed as described (Grosveld et al., 1987). RNA analysis RNA was isolated from transfected cell populations and 13.5 day fetal mouse liver by the method of Auffray and Rougeon (1980). Quantitative SI nuclease analysis using human 13-globin and mouse caglobin DNA probes was performed as described (Kollias et al., 1987; Antoniou et al., 1988; Talbot et al., 1989). DNA analysis Southern blotting was performed essentially as described by Southern (1975) using nylon membranes (Nytran, Schleicher and Schiull) and the hybridization conditions described by Church and Gilbert (1985). Copy numbers of the human (3-globin gene were determined by laser densitometry using the signal of the endogenous mouse Thy-I gene as an internal loading control. 13.5 day transgenic fetuses were screened for mosaicism of the transgene by comparing the human /3-globin/mouse Thy-I ratio in DNA from placenta, body, head and yolk sac. DNase I footprinting The 225 bp HphI-Fnu4HI fragment of HSS 2 was blunted and cloned in both orientations in the SmaI site of M13mp8. 1 1tg of single-stranded template DNA was annealed to 5 pmol of kinased Universal 17 mer sequencing primer and extended using Klenow polymerase and all four cold dNTPs. The reaction products were phenol extracted and passed over a Sephadex G50 column. After overnight digestion with 10 U EcoRI, the DNA was phenol extracted, ethanol precipitated and dissolved in 200 A1 of TE buffer. Each footprinting reaction contained 1 A1 of the probe mentioned above (3000 c.p.m.), 2 jg poly(dI-dC) and 10-50 Ag nuclear protein as described by Wall et al., 1988, but was pre-incubated for 10 min on ice. The samples were analysed on 6% sequencing gels, using dideoxy sequencing reactions as markers. Nuclear extracts were prepared according to Gorski et al. (1986), with modifications as described in Wall et al. (1988). HeLa nuclear extracts were prepared according to Dignam et al. (1983).
Gel mobility shift assays and competition studies Gel mobility shift assays were performed as described previously (deBoer et al., 1988), using 5 lAg MEL nuclear extract per reaction. Competitions were done by adding 100-fold molar excess of the indicated double-stranded oligonucleotides before addition of the extract. Nucleotide sequences of the oligonucleotides used in these studies were:
oligonucleotide 2 sense strand: ATCACAGGTTCCAGGGAGGGTGGGGTGGGGTCAGGGCTGGCCAC
2166
oligonucleotide 4 sense strand: GCTCAGATAGGTGGTTAGGTCAGGTTGGTGGTGCTGGGTGGAGTCCATGACTCCCAG oligonucleotide 6 sense strand: TAGAGGGGAGACATGGGAAAGGTGGGGGAGGCACAGCATAG Spl oligonucleotide late strand: CGATGGGCGGAGTTAGGGGCGGGACTAT CACC oligonucleotide sense strand: CGATCCGTAGAGCCACACCCTAGGTAT
Acknowledgements We are grateful to E.deBoer for assistance and to Cora O'Carroll for preparation of the manuscript. S.P. was supported by an EMBO long-term fellowship, D.T. by an NSERC (Canada) and a Commonwealth scholarship and P.F. by an NIH postdoctoral fellowship. This work was supported by the Medical Research Council (UK).
References Antoniou,M., deBoer,E., Habets,G. and Grosveld,F. (1988) EMBO J., 7, 377-384. Auffray,C. and Rougeon,F. (1980) Eur. J. Biochem., 107, 303-314. Behringer,R.R., Hammer,R.E., Brinster,R.L., Palmiter,R.D. and Townes,T.M. (1987) Proc. Natl. Acad. Sci. USA, 84, 7056-7060. Blom van Assendelft,M., Hanscombe,O., Grosveld,F. and Greaves,D.R. (1989) Cell, 56, 969-977. Chada,K., Magram,J. and Costantini,F. (1985) Nature, 319, 685-688. Church,J.M. and Gilbert,W. (1984) Proc. Natl. Acad. Sci. USA, 811, 1991-1995. Collins,F.S. and Weissman,S.M. (1984) Prog. Nucleic Acid Res. Mol. Biol., 31, 315-462. Collis,P., Antoniou,M. and Grosveld,F. (1990) EMBO J., 9, 233-240. Curtin,P.T., Liu,D., Liu,W., Chang,J.C. and Kan,Y.W. (1989) Proc. Natl. Acad. Sci. USA, 86, 7082-7086. deBoer,E., Antoniou,M., Mignotte,V., Wall,L. and Grosveld,F. (1988) EMBO J., 7, 4203-4212. Dignam,J., Lebowitz,R. and Roeder,R. (1983) Nucleic Acids Res., 11, 1475-1489. Dynan,W. (1989) Cell, 58, 1-7. Evans,T. and Felsenfeld,G. (1989) Cell, 58, 877-885. Evans,T., Reitman,M. and Felsenfeld,G. (1988) Proc. Natl. Acad. Sci. USA, 85, 5976-5980. Forrester,W.C., Takegawa,S., Papayannopoulou,T., Stamatoyannopoulos,G. and Groudine,M. (1987) Nucleic Acids Res., 15, 10159-10177. Forrester,W.C., Novak,U., Gelinas,R. and Groudine,M. (1989) Proc. Natl. Acad. Sci. USA, 86, 5439-5443. Fraser,P., Hurst,J., Collis,P. and Grosveld,F. (1990) Nucleic Acids Res., in press.
Gidoni,D., Kadonaga,J.T., Barrera-Saldana,H., Takahashi,K., Chambon,P. and Tjian,R. (1985) Science, 230, 511-514. Gorski,K., Carneiro,M. and Schibler,U. (1986) Cell, 47, 767-776. Greaves,D.R., Fraser,P., Vidal,M., Hedges,M., Ropers,D., Luzzatto,L. and Grosveld,F. (1990) Nature, 343, 183-185. Grosveld,F., Blom van Assendelft,G., Greaves,D.R. and Kollias,G. (1987) Cell, 51, 975-985. Gustafson,T.A. and Kedes,L. (1989) Mol. Cell. Biol., 9, 3269-3283. Jackson,S. and Tjian,R. (1988) Cell, 55, 125-133. Kadonaga,J.T., Carner,K.R., Masiara,F.R. and Tjian,R. (1987) Cell, 51, 1079-1080.
Kollias,G., Hurst,J., deBoer,E. and Grosveld,F. (1987) Nucleic Acids Res., 15, 5739-5747. Magram,J., Chada,K. and Costantini,F. (1985) Nature, 315, 338-340. Maniatis,T., Goodbourn,S. and Fischer,J.A. (1987) Science, 236, 1237-1245. Mignotte,V., Eleouet,J.F., Raich,N. and Romeo,P.-H. (1989a) Proc. Natl. Acad. Sci. USA, 86, 6545-6552. Mignotte,V., Wall,L., deBoer,E., Grosveld,F. and Romeo,P.-H. (1989b) Nucleic Acids Res., 17, 37-54. Ofverstedt,L.-G., Hammarshom,K., Balgobin,N., Kjerten,S., Petterson,U. and Chattopadhyaga,J. (1984) Biochim. Biophys. Acta., 732, 120-126. Ryan,T.M., Behringer,R.R., Townes,T.M., Palmiter,R.D. and Brinster,R.L. (1989) Proc. Natl. Acad. Sci. UJSA, 86, 37-41.
,B-globin dominant control region
Schule,R., Muller,M., Kaltschmidt,C. and Renkawitz,R. (1988) Science, 242, 1418-1420. Serfling,E., Jasin,M. and Schaffner,W. (1985) Trends Genet., 1, 224-230. Southern,E.M. (1975) J. Mol. Biol., 98, 503-5 17. Talbot,D., Collis,P., Antoniou,M., Vidal,M., Grosveld,F. and Greaves,D.R. (1989) Nature, 338, 352-355. Talbot,D., Philipsen,S., Fraser,P. and Grosveld,F. (1990) EMBO J., 9, 2169-2178. Tsai,S.-F., Martin,D.I.K., Zon,L.I., D'Andrea,A.D., Wong,G.G. and Orkin,S.H. (1989) Nature, 339, 446-451. Tuan,D., Solomon,W., Qiliang,L.S. and Irving,M.L. (1985) Proc. Natl. Acad. Sci. USA, 32, 6384-6388. Tuan,D.Y.H., Solomon,W.B., London,I.M. and Lee,D.P. (1989) Proc. Natl. Acad. Sci. USA, 86, 2554-2558. Wall,L., deBoer,E. and Grosveld,F. (1988) Genes Dev., 2, 1089-1100. Xiao,J.-H., Davidson,J., Macchi,M., Rosales,K., Vigneron,M., Staub,A. and Chambon,P. (1987) Genes Dev., 1, 794-807. Received on January 29, 1990; revised on March 27, 1990
2167
